Significant figures and rounding criteria: How not to exaggerate calculation results

Just because a calculator displays a result of 69.104 does not necessarily mean the measurement result is known to three decimal places. The extent to which the input value has been measured and the purpose for which the result will be used influence the determination of the display digits. Significant figures and rounding are methods to adjust the impression of a calculation rather than rules for making numbers look pretty. All figures in this article are hypothetical data for conceptual explanation purposes and are not the results of actual material tests or precision measurements.

1. Significant figures and decimal places are different standards.

Decimal places refer to the number of digits used after the decimal point. Significant figures are the number of digits retained to express the magnitude of a number. For example, 0.00450 is written with five decimal places, but in standard measurement notation, it is read as three significant figures by counting the 4th, 5th, and the final 0. The leading zeros serve to indicate the position of the decimal point. If the final 0 is omitted to write 0.0045, the displayed significant figures are reduced to two.

Notations ending in zeros, such as 4500, can be ambiguous without context. Writing 4.5 × 10³ reveals the intention of two digits, while writing 4.50 × 10³ reveals the intention of three digits. However, if the documentation separately defines it as "a number rounded to the nearest hundred," that explanation must be read as well. It is important not to estimate equipment accuracy or statistical uncertainty based solely on the appearance of the number.

Notation Example	Readable indication intent	Separately required information
0.00450 m	usually three significant figures	Measurement Method · Uncertainty
4.50 × 10³ pieces	Quantity displayed in three digits	Whether it is an actual coefficient or an estimate
12.30 mm	Display up to two decimal places	Display resolution and accuracy
69 mm²	Area displayed to integer decimal places	Input values and rounding criteria
34.57 ± 0.12 units	Example of aligning the last digits of the value and uncertainty	Types of Uncertainty and Calculation Conditions

2. Distinguish between measured values and precisely defined numbers.

The role assigned to numbers differs between a count determined by actual measurement, such as '3 samples,' and a length of 3.0 mm read by equipment. Multiplying by the exact count of 3 in a calculation does not automatically reduce the measured value to a single significant figure. Conversion relationships defined by unit definitions must also be distinguished from values obtained through measurement. Changing units using the relationship that 1 cm equals 10 mm does not constitute obtaining new measurement information.

For example, converting an explanatory length of 12.3 mm to cm results in 1.23 cm. Although the number of decimal places increases due to the change in unit, the same physical quantity and display information have been maintained. Writing 1.230000 cm does not increase the basis for the measurement. Conversely, writing 1 cm for the sake of readability significantly reduces the information. When determining the number of decimal places after unit conversion, consider the information retained in the original value and the purpose of use, rather than the number of decimal places.

3. Understand the textbook rules of addition and multiplication.

In basic classes, the rule of fitting addition and subtraction to the roughest decimal places, and multiplication and division to the fewest significant figures, is frequently used. This rule is a convenient starting point for practicing the display of calculation results. It is not a universal law that accurately conveys uncertainty in all actual measurements.NIST numerical reporting dataalso explains that it is difficult to handle the arithmetic of measured values using only uniform digit rules.

The virtual addition 12.34 + 0.6 results in a calculated value of 12.94. In a lesson example applying the decimal place rule, the result is written as 12.9 because 0.6 is displayed to the first decimal place. Multiplying the length of the virtual rectangle, 12.34 mm, by its width, 5.6 mm, gives 69.104 mm². In an exercise applying the multiplication rule, this can be displayed as 69 mm² to accommodate two significant figures for the width. This is not the result of actually calculating the area uncertainty.

If you do not know why the input value was written only to that specific decimal place, there are limitations to applying the rules. Even with the same number 5.6, the interpretation varies depending on whether it is a value read from the equipment, a value that has already been rounded to the average, or an exact ratio defined in the design. In reports, please include the source of the input value along with the calculation formula. Distinguishing between the notation rules for problem solving and uncertainty calculations in experimental reports can reduce the error of judging reliability based solely on the number of decimal places.

4. Keep the intermediate results and round at the final step.

In sequential calculations, rounding the value at each step to the nearest whole number can alter the result. For illustrative purposes, let's divide 1 by 3 and then multiply it by 3 again. If we reduce the intermediate value to 0.33, the result becomes 0.99. However, if we display the result at the end while maintaining a sufficient number of decimal places, we obtain a result closer to the original 1. This is not a method to eliminate actual measurement errors, but rather a method to reduce unnecessary loss of computational information.

In spreadsheets, you must distinguish between changing a cell display to two decimal places and converting the calculated value itself using a rounding function. Although it appears as 0.33 on the screen, the internal value may retain more decimal places. When copying and passing text, situations may arise where only the displayed value is transmitted. If someone needs to recalculate the result, provide the original input, the formula, and the final display rules together. It is better not to attempt to reconstruct the calculation process based solely on the value visible on the screen.

Reproducible calculation records are stored by separating the raw data, the values used in the calculation, and the values for reporting. The raw data is left as received, values with verified units and processing conditions are used during calculation, and the number of decimal places is reduced at the reporting stage. For example, separating the inputs 12.34 and 5.6, the internal calculation 69.104, and the reported value 69 allows you to identify where rounding occurred. Situations where the reported value is used again as raw data for the next calculation should be avoided, or such limitations should be disclosed.

5. Establish the rule for the case where it is exactly half.

We are usually taught to 'round up to 5 or higher,' but there are several ways to handle values that are exactly in the middle.NIST SI Guideexplains rounding that makes the last digit to be kept an even number when the first digit to be discarded is exactly 5 and the rest are all zeros. Other tasks or tools may use a rounding-up method. Whichever method is used, it is important to use it consistently throughout the document and to check the rules of the tool being used.

The difference becomes apparent when looking at an explanatory example of rounding the positive number 2.345 to two decimal places. For an exact decimal input, rounding up results in 2.35, while the half-even method results in 2.34. 2.355 becomes 2.36 in both methods. You must first determine whether the part to be discarded is exactly half or greater than half. Since 2.3451 is greater than exactly half, it becomes 2.35 in the even method as well. Do not look only at the single digit 5 and ignore the digits following it.

This difference does not mean the data is incorrect, but rather that the rounding policies are different. When the two reports differ only in the last digit, compare the raw data, units, rounding step, and halving rule in that order. If the data includes negative numbers, you must also distinguish whether "rounding up" refers to a mathematical ceiling function or halving in the direction of the absolute value. The numerical comparisons in this article are limited to positive decimal examples to illustrate the concept.

6. Clearly express decimal intent in the code.

Computer binary floating-point numbers cannot accurately store some decimal fractions. Just because you enter the number 2.345 does not mean the internal value is always the exact decimal 2.345. When explaining rounding comparisons, the method of constructing decimal values in strings clarifies the intent. The following Python example is a calculation for illustrative purposes and is not a test that guarantees the accuracy of measurement data or the actual results of a specific program.

from decimal import Decimal, ROUND_HALF_UP, ROUND_HALF_EVEN

value = Decimal("2.345")
step = Decimal("0.01")
print(value.quantize(step, rounding=ROUND_HALF_UP))    # 2.35
print(value.quantize(step, rounding=ROUND_HALF_EVEN))  # 2.34

The key is to convert numbers into strings and specify the target decimal places and rules. Converting a value already generated from floating-point operations into a string does not automatically make the previous calculation process accurate. In your data processing code, be sure to clearly distinguish between input conversion, calculation precision, and final rounding. You can determine whether precise decimal calculations are required or if approximate calculations are sufficient based on the purpose of use.

7. If there is uncertainty, align the value and the last digit together.

When reporting uncertainty in measurement results, the meaning of the uncertainty is more important than the simple significant figures rule. For illustrative purposes, let us assume a calculated value of a quantity is 34.567 and its standard uncertainty is 0.123. If you choose to report the uncertainty as 0.12 (two significant figures), you might consider configuring the value to be 34.57, also including two decimal places. The uncertainty here is not a value obtained from actual repeated measurements or equipment specifications.

Simply adding '±' does not complete the report. You must explain whether it is standard uncertainty or expanded uncertainty, what inclusion factor was used if expanded, and what the measurement conditions and units are. This article explains the principles of decimal place display and does not serve as a substitute for the uncertainty estimate for a specific test. It also avoids assuming that the error is exactly 0.001 simply because the equipment display shows up to the third decimal place. Display resolution and the total measurement uncertainty are not the same concept.

8. Items to check before reporting

• Did you distinguish between the measured value and the exact count/defined transformation relationship?
• Did you confuse the number of significant figures with the number of decimal places?
• Did you record the unit and source of the input value together?
• Did you reduce the intermediate calculation value too early?
• Did you specify the exact half-processing rule?
• Have you checked the difference between the display format and the internally calculated value?
• If you actually calculated the uncertainty, did you write down its type and conditions?
• Did you store raw data and reporting values separately?

A result with more digits is not always a more accurate result. Conversely, removing all necessary digits results in the loss of information useful for calculations. First, define the meaning of the input and the purpose of the report, retain the information during the calculation, and apply the rules in the final display. The habit of including the unit, basis, and processing method along with the numbers helps convey calculation results without exaggeration.

Official Sources and Writing Standards

• NIST SI Guide: Unit Conversion and Rounding
• NIST Special Publication 747: Processing and Reporting of Measurement Data
• Python Official Documentation: Decimal and Rounding Methods

Data Verification Date: 2026-10-10. This explanation was generated by AI based on actual official data. Calculations, codes, and verification examples marked separately are for illustrative purposes only and are not the results of direct testing or actual measurements of the user environment. We will re-verify whether there have been any changes to the functions and data on the publication date.

The process of transferring calculated values to report values

1. Check the meaning and units of the input values

→

2. Distinguishing between measured values and exact numbers

→

3. Maintain sufficient digits for the median

→

4. Determine target placeholder and half processing rules

→

5. Report final value along with uncertainty and conditions

This is an illustration for illustrative purposes only and is not actual product screen or measurement data.